Psychophysical Colour

11.15 Psychophysical Colour

This is a recording of my presentation "Psychophysical Colour", given at AIC2022 Sensing Colour, the virtual Midterm Meeting of the International Colour Assocaiation held in Toronto on June 13-16, 2022.


[00:00] The CIE International Lighting Vocabulary (ILV) defines the word “colour” in two distinct senses, “perceived” colour and “psychophysical” colour, yet the concept of psychophysical or “colorimetric colour” is regarded by many colour specialists as philosophically suspect or even nonsensical. How can a colorimetric specification be a colour?  

[00:25] “Colour” in the perceptual sense or “perceived colour” is defined as a “characteristic of visual perception” that can be described by combinations of six attributes, each of which is in turn defined, either directly or indirectly, as an “attribute of a visual perception”.

“Colour” in the psychophysical sense is defined as a “specification of a colour stimulus” in terms of values such as three tristimulus values”, which are in turn defined as the “amounts of the reference colour stimuli required to match the colour of the stimulus considered”, giving as examples R, G, and B, and X, Y, and Z in CIE colorimetry. When we speak of “colour measurement”, “colour difference formulae”, many “colour spaces” and the 16.7 million RGB “colours” on our screens, we are using the word “colour”, some would say improperly, in this second sense.

This presentation will review the connections between colour stimuli, colour perceptions, and colorimetric specifications in order to defend the view that both these things can be considered to be colours.

Colours of lights

[01:36] Newton showed that the colour of an isolated light can be predicted from the overall balance of its spectral components considered in relation to a two-dimensional circuit of directions of imbalance relative to light perceived as white, such as daylight. The hue of the light can be predicted from the direction of imbalance, and the saturation or “distance from whiteness” can be predicted from the amount of imbalance. Another way of saying this is that the perceived colour of an isolated light is the way in which we perceive the overall balance of its spectral components. Thus, white as a colour of an isolated light is the way in which we perceive an overall spectral balance similar to that of daylight, and whitish orange as a colour of an isolated light is the way in which we perceive a spectral distribution biased in a certain way towards the longer wavelengths of the spectrum.

[02:34] Of course, our colour perceptions are not based on instrumental measurements but on the responses of the human visual system and are therefore shaped in part by the characteristics of that system. We now know that Newton’s directions of perceivable imbalance form a two-dimensional circuit because they arise from the responses of three cone cell types. Comparison of these responses by the process of cone opponency allows a colour-normal observer to detect a two-dimensional circuit of directions of spectral imbalance towards long, middle, short, or long and short wavelengths respectively.

[03:15] Our dependence on three receptor types means that physically different lights will match in perceived colour if they evoke the same relative response of the three cone cell types. For example, the very different spectral distributions of the three lights shown here all have an even overall balance, lacking any overall bias towards long-, middle- or short-wavelengths from the viewpoint of the human visual system. This overall balance is the perceivable property that these physically different but visually matching lights, called metamers, have in common.

[03:51] Perceived colours can be influenced by a variety of factors in addition to the spectral properties of the stimulus, as acknowledged in this note appended to the CIE definition of “perceived colour”. Nevertheless, despite the importance of these other factors, it remains reasonable to say that in many ordinary circumstances we perceive variations in the spectral composition of light at the level of its long-, middle- and short-wavelength components as different colours. This of course is why we’re all looking at machines that work by emitting long-, middle- and short-wavelength light in different proportions.

Colours of objects

[04:30] If we can freely examine an object in daylight, the colour we perceive the object as having is usually a very good indication of its overall spectral reflectance at the level of its long-, middle- and short-wavelength components; or to put this another way, under such conditions an object’s perceived colour is the way in which we perceive this overall spectral reflectance. The hue and chroma are good indications of respectively the direction and amount of this imbalance, and the lightness indicates the proportion of light that the object reflects. Here again, though, our perceptions are shaped by the workings of the human visual system, because the response of our visual system tapers towards the extremes of the spectrum, and so wavelengths near these extremes have a relatively weak influence on our perception. Thus, our perception of the amount of light that an object reflects relates not directly to its physical energy but to its physical energy weighted by the wavelength-by-wavelength response of the human visual system, called its relative luminance.

[05:36] I should mention that under some circumstances even freely examined objects can evoke perceived colours that are poorly correlated with their spectral properties, notably when the object is an image area in a depiction of an illuminated scene, as in the wonderful colour constancy illusions of Purves and Lotto. In viewing these images the perceived colours of the virtual objects depicted in the scene are so visually insistent that it can be extremely difficult to attend to colour perceptions relating to the image areas themselves, unless we break the representational spell of the image areas by providing a target seen as being outside the depicted illumination, when suddenly we can make veridical comparisons, as I’ve discussed elsewhere*.

[06:33] Our ability to perceive the overall spectral reflectance of an object as its colour is most effective in daylight or similar illumination. Under spectrally biased illumination our visual system has the capacity, to a degree, to adapt to the spectral bias, making the illumination appear less strongly coloured than it would otherwise, and quite separately to this, to a degree, to disentangle or parse a scene into colours relating to illumination and spectral reflectance, so that we might perceive some things in this reddish scene as white and grey objects under reddish illumination. Nevertheless, our capacity to “see” spectral reflectance as colour diminishes as the spectral bias of the light increases, and under monochromatic illumination we perceive only variations in the object’s reflectance of the single wavelength present.  

Colorimetry of lights

[07:33] Colorimetric specifications are designed to ignore physical differences that are not perceivable to a human observer, and to record just those differences that we perceive as colour differences. For example, position in a chromaticity diagram such as the CIE 1931 diagram represents the overall balance of wavelengths in a light, at the level of the long-, middle- and short-wavelength components, as detected by the human visual system, given the necessary assumption of a “standard” human observer. The three very different spectral distributions here have the same such balance, lacking any overall bias at this level, and so plot at the same point.

A colorimetric specification of a light represents a class of physically varied lights having a common disposition to evoke a perceived colour, in the sense that they would be expected to match in perceived colour when viewed under the same conditions. A colorimetric specification of a light may thus be said to quantify what Newton called colours “in the Rays”, meaning a light’s “Power and Disposition to stir up a Sensation of this or that Colour”. Note however that a colorimetric specification does not correspond to a single “inherent” perceived colour; our three lights will match in colour when viewed in the same setting, but viewed in different settings the common perceived colour of these matching lights would vary.

Colorimetry of objects

[09:07] As with lights, colorimetric specification of objects is intended to ignore differences that we do not perceive as colour differences, but in this case of spectral reflectance . For example, each Munsell chip is manufactured to embody a colorimetric specification of the chromaticity (xy) and the relative luminance (capital Y) of the light that the chip would reflect under Illuminant C. For a spectrally neutral object the chromaticity is the same as that of the illuminant, and other chromaticities indicate various directions and degrees of spectral imbalance. Each xyY value represents a class of physical spectral reflectances that are metameric (that is, match) under Illuminant C for the standard observer. Granted these necessary assumptions of a standard observer and a specific daylight illuminant, CIE xyY values quantify for practical purposes the perceivable property that we perceive as the colour of the chip in daylight.

Colour ontology

[10:17] In philosophy of colour there are many different theories of colour ontology, which is concerned with questions about the fundamental nature of colour. These theories differ among themselves in part over what the word “colour” is taken to apply to, for example to colour perceptions such as red, blue etc., or to individual spectral reflectances or distributions, or to the power or disposition of lights and objects to cause a perception of red, blue etc., or to cause such a perception in a given perceiver and environment.

[10:55] In defining two senses of the word “colour” the CIE International Lighting Vocabulary in effect expresses a pluralist ontology that acknowledges that we may wish to use the word “colour” either for our perceptions of colour or for the perceivable properties of lights and objects that these perceptions are based on. “Psychophysical colour” refers to the colorimetric measures devised to specify the perceiver-dependent property of a light, or of an object viewed in daylight, that determines its disposition to appear this or that colour. In terms of Hardin’s well-known quote – “Colored objects are illusions, but not unfounded illusions” - “psychophysical colour” specifications identify for practical purposes the properties of lights and objects that our “illusions” of perceived colour are founded on.

*Colour Constancy Illusions and Painting,


I am very grateful to Robert Hirschler, Barry Maund, Mohan Matthen, and Derek Brown for generously providing comments on an earlier draft of this manuscript. All of their comments led to corrections or clarifications, but responsibility for any remaining errors is entirely my own.

This page published June 16, 2022.
< Index of works